Back to skill
Skillv1.0.0
ClawScan security
arxiv-to-obsidian · ClawHub's context-aware review of the artifact, metadata, and declared behavior.
Scanner verdict
BenignMar 20, 2026, 7:19 AM
- Verdict
- benign
- Confidence
- high
- Model
- gpt-5-mini
- Summary
- The skill's code, scripts, and instructions are coherent with its stated purpose (fetch arXiv RSS, translate titles/abstracts, and write via Obsidian CLI); no obvious misdirection or hidden endpoints were found, but it will send paper text to the 'claude' CLI/service for translation and requires trusting the obsidian/claude binaries on your machine.
- Guidance
- This skill appears to do what it says: fetch arXiv RSS, translate titles/abstracts, and write a Markdown table to an Obsidian note via the Obsidian CLI. Before installing/using it: - Inspect and set scripts/config.sh (VAULT_NAME, VAULT_FOLDER, NOTE_NAME) to values you control; do not run with defaults if they don't match your vault. The README emphasises this for a reason. - Run a dry run first (DRY_RUN=1 ./fetch-arxiv.sh) to verify the output and the target vault/folder/note are correct. - Understand that translations are performed by the 'claude' CLI: your paper titles and abstracts will be sent to whatever backend 'claude' is configured to use. If those texts are sensitive, do not use this skill or configure 'claude' appropriately. - Verify you trust the local obsidian and claude binaries. The skill delegates file creation/deletion to the Obsidian CLI; ensure the CLI behaves as you expect and that VAULT_FOLDER values cannot be used to create files outside the intended vault in your environment. If you want tighter privacy, replace the translator to use an on-device translator or a service you control, or remove/modify the call to 'claude'.
Review Dimensions
- Purpose & Capability
- okName/description match the included scripts and README: the skill fetches arXiv RSS, parses up to N papers, translates titles/abstracts, formats a Markdown table, and writes/appends via the Obsidian CLI. Required tools (curl, python3, obsidian, claude) are reasonable for these actions.
- Instruction Scope
- noteSKILL.md clearly limits scope to fetching RSS, parsing metadata, translating, and writing via Obsidian CLI. The runtime scripts follow that scope and do not read unrelated system files. Important operational behavior: all paper text (title+abstract) is passed to the 'claude' CLI for translation — that transmits data to whatever backend the 'claude' tool is configured to use.
- Install Mechanism
- okNo install spec is provided (instruction-only with shipped scripts). All code is included in the repo; nothing is downloaded or extracted at install time. This is low-risk from an install/download perspective.
- Credentials
- noteThe skill requests no environment variables explicitly, instead using a local scripts/config.sh for VAULT_NAME/VAULT_FOLDER/NOTE_NAME. It requires 'claude' and 'obsidian' binaries; 'claude' typically requires credentials or a configured client and will forward content externally — this is proportionate to translation but has privacy implications. The skill doesn’t request unrelated credentials or access and does not embed secret-handling code, but you should be aware that translation occurs via the installed 'claude' tool.
- Persistence & Privilege
- okalways:false and no persistent installation behavior is requested. The skill uses the Obsidian CLI to create/append notes inside a vault; it does not modify other skills or global config. No autonomous 'always' presence or elevated privileges are requested.
